Cytosine DNA methylation is an evolutionarily conserved epigenetic mark that plays critical roles in diverse biological processes, including gene and transposon silencing and imprinting. In mammals, DNA methylation mostly occurs in the symmetric dinucleotide CG sites. In the model plant Arabidopsis thaliana, DNA methylation frequently occurs at cytosine bases in all sequence contexts (CG, CHG and CHH, where H represents A, C or T).
In Arabidopsis, de novo DNA methylation is established by a process known as RNA-directed DNA methylation (RdDM). RdDM in plants not only requires the upstream production of 24-nucleotide (nt) small interfering RNAs (siRNAs) and the downstream recruitment of de novo DNA methyltransferase DOMAINS REARRANGED METHYLTRANSFERASE 2 (DRM2), the production of RNA polymerase V (Pol V)-dependent intergenic non-coding (IGN) transcripts plays a crucial role likely in serving as scaffolds for siRNAs binding. Pol V is required for DNA methylation and gene silencing and has been shown to be transcriptionally active in vitro. However, the characteristics of Pol V transcripts is poorly understood probably due to its low abundance. In this dissertation, to better understand the features of Pol V transcripts in vivo, I will first describe the application of a technique modified from global nuclear run-on (GRO) assay to characterize nascent Pol V transcripts at genome-wide level. With this technique, we captured Pol V nascent transcripts and we uncovered a novel mechanism of ARGONAUTE4/6/9 (AGO4/6/9) dependent, small-RNA-guided co-transcriptional slicing of nascent Pol V transcripts.
With the fast development of genome editing recent years, epigenetic modification including targeted DNA methylation and demethylation also becomes attractive for its capability of stably regulate gene expression. In order to develop site-specific and efficient tools for DNA methylation targeting, we tethered artificial zinc finger protein recognizing specific DNA sequence to various RdDM proteins (ZF-RdDM) in Arabidopsis. With this tool, we studied the hierarchy of action within RdDM pathway by testing their ability to target methylation in different mutant backgrounds. Also, at thousands of ZF-RdDM off target sites, we characterized the ectopic siRNAs production, Pol V recruitment and DNA methylation establishment and found that simultaneously recruiting both arms of the RdDM pathway, siRNA biogenesis and Pol V recruitment, dramatically enhanced targeted methylation. We then also developed a tool to target DNA demethylation in plants by fusing the catalytic domain of the human demethylase TEN-ELEVEN TRANSLOCATION1 (TET1cd) and an artificial zinc finger protein or CRISPR/dCas9 system.
Finally, I will discuss DNA methylation landscape in human embryonic stem cells (hESCs). hESCs are morphologically and transcriptionally similar to stem cells derived from the mouse post-implantation epiblast. Thus, hESCs are typically considered to exhibit ‘primed’ pluripotency. Various culture conditions have been developed to promote maintenance and self-renewal of hypomethylated ‘naive’ hESCs. We have discovered that reverting primed hESCs to naive hESCs results in a Stage Specific Embryonic Antigen 4 (SSEA4)-negative population with a transcriptional program resembling the human pre-implantation epiblast. However, we also discovered that the methylation landscape of naive hESCs in vitro is distinct from human epiblast in vivo with a lost ‘memory’ of methylation state at primary imprints and human oocyte.